Zhiwei Liu would like to dedicate this paper to the honor of Professor Kumar Krishna for his friendship, kindness, professional encouragement, and the good times at the AMNH.

Catantopidae (Acridoidea, Orthoptera) is a well-known grasshopper family; its members include some of the most notorious pests in agriculture, including Schistocerca gregaria (Forsköl), Oxya spp, and Melanoplus spp (Hill 1987). The family is by far the largest and the most diverse acridoid family, consisting of over 3000 species in about 640 genera mainly distributed in the tropical and subtropical areas of the world (Vickery and Kevan 1983).

The previous classifications of Acridoidea (Orthoptera) have been predominantly utilitarian; existing classifications of the superfamily almost entirely ignored phylogenetic relationships among taxa. Among the various classification systems or schemes of acridoids (Dirsh 1961, 1975, Harz 1975, Otte 1981, Yin 1982, Xia 1994, Li and Xia 2002) and several other classifications specifically proposed for the Catantopidae (Tinkham 1940, Mistshenko 1952, Harz 1975), there exist a great deal of disagreement concerning the classification within the family (Table 1), which cannot be easily settled because of the lack of phylogenetic studies. The most influential classification systems of Acridoidea at the present are still the one established by Dirsh (1956) and its modified versions (Dirsh 1961, 1975). The classifications by Dirsh are based on extensive comparative studies of the genitalia morphology of both sexes as well as other morphological characters, emphasizing especially the importance of the morphology of phallic complex and epiphallus in defining higher taxa. Several other authors also proposed their own classification for the Acridoidea (Rehn and Grant 1960, Uvarov 1966, Jago 1971, Vickery and Kevan 1983, Liu 1991). Otte (1981, 1984) adopted a compromised version of the various systems in his monographic treatment of North American grasshoppers. These classifications, although different, have one thing in common: all are entirely based on overall similarity and make little, if any, reference to phylogenetic relationship.

The need for a classification of the grasshoppers and locusts based on phylogeny, rather than based on overall similarity, is obvious. Yin (1982) pointed out the importance of distinguishing between plesiomorphic and derived features in the classification of the acridoids and paid special attention to the transformation series of antennae, wings, and stridulating apparatuses and tympanum. However, Yin’s classification of Acridoidea based on his studies of the Chinese members of the group was not based on phylogenetic analysis and his circumscriptions of higher-rank taxa were often based on characters that obviously have been obtained through convergent evolution. Key and Colless (1993) attempted to conduct a cladistic (and phenetic) analysis of the subfamily Catantopinae from Australia. They coded 104 male external characters for 166 genera and conducted a series of analyses from typical phenetic approaches to cladistic methods as implemented in PAUP (version not mentioned).The results of this particular study showed almost no resolution of relationships or useful clustering except for a few ‘low-level groups’. The authors consequently did not even bother to present the cladograms and resolved to ‘develop a classification by traditional non-quantitative methods’.

There has been an increased interest in recent years in the phylogenetic relationship of orthopteroid insects in general (Flook and Rowell 1997a, 1997b, 1998, 1999; Flook et al. 2000, Rowell and Flook 1998, Yin et al. 2003) and the acridoids in particular (Amedegnato et al. 2003, Chapco et al. 2001, Litzenberger and Chapco 2001, 2003, Ren et al. 2002, Xi and Zheng 1997, Xu et al. 2003, Xu and Zheng 1999, 2002; Zheng and Qiao 1998). Most of these recent studies are based on molecular data with relatively limited taxon sampling; the few morphology-based studies either targeted at lower level relationship (e.g., within genus, Xu et al. 2003, but see Song and Wenzel 2007) or are characterized by sporadic taxon sampling (Ren et al. 2002, Xu and Zheng 1999, 2002; Zheng and Qiao 1998). Therefore, the potential of morphology in resolving higher-level phylogeny within Orthoptera and Acridoidea has not been fully explored.

This lack of higher-level phylogenetic study of Catantopidae leads to a lack of stability in the classification within the family (Table 1), which is unusual for such a well-known and economically important group. In this paper, we present the first comprehensive phylogenetic analysis of the family Catantopidae based on morphology by sampling extensively from the Chinese fauna. Our purpose is to (1) conduct an exploratory phylogenetic analysis of the phylogenetic relationship within the family represented by the Chinese members, (2) provide an objective evaluation of the usefulness of morphological characters in phylogeny reconstruction of the acridoids in general and the catantopids in particular, and (3) provide a general framework for taxon sampling in future studies of acridoid phylogeny on a global basis.

The name Catantopidae, or its original form Catantopinae as subfamily, has had a long history of divergent usages (Key and Colles 1993). The modern definition of Catantopidae took after the name of Cyrtacanthacrinae (Tinkham 1940, Roberts 1941) and was subsequently assigned subfamily status as Catantopinae by Mistshenko (1952). The latter author further assigned the members of the subfamily in the fauna of the former Soviet Union and adjacent countries into thirteen tribes, and considered Egnatiinae as a separate group from the Catantopinae. Mistshenko (1952) was mostly accepted by later authors, including Dirsh (1961), Uvarov (1966), and Harz (1975). Dirsh (1975) later divided Mistshenko’s Catantopinae into two families, Hemiacrididae and Catantopidae, and included Egnatiinae in the family Catantopidae. Yin (1982) also divided Mistshenko’s Catantopinae into two families, Acrididae and Oedipodidae, but treated Egnatiidae as a separate family. Xia (1994) included in the family Catantopidae some of the subfamilies of Oedipodidae in Yin’s system and raised most of the tribes in Mistshenko’s system to subfamilies. The Xia System has been adopted for the recent monographic treatment of the Chinese fauna of Catantopidae (Li and Xia 2002). The classification scheme used by Otte (1981, 1984) in his monographic treatments of the grasshoppers of North America north of the Gulf of Panama, although unexplained, is obviously completely utilitarian without reference to phylogenetic relationship among groups. The Otte classification was later expanded and adopted by the author in his multi-volume catalog of the orthopteran insects of the world (Otte 1994a, 1994b, 1994c, 1995a, 1995b), which in turn has been eventually published as a searchable online database, the Orthoptera Species File (Eades et al 2011). While the Orthoptera Species File database is tremendously useful for taxonomic purposes, species groups defined by earlier classification systems were often used in phylogenetic studies on Acrididae / Acridoidea at levels of tribe and above (Litzenberger and Chapco 2003; Song and Wenzel 2007). A comparison of catantopid classification systems by various authors is given in Table 1.

Catantopidae in our view is readily defined by the unmistakable synapomorphy of having a distinct prosternal process between the forecoxae. Although some species of Pamphagidae and Pyrgomorphidae have a lamellate process on the prosternum, the process in these species is on the anterior margin of the prosternum and is obviously an independently evolved feature not homologous to the prosternal process between the forecoxae observed in Catantopidae. Nonetheless, as shown in Table 1, there was considerable disagreement among earlier authors about the definition of Catantopidae, which obviously arose from the fact that earlier acridologists defined higher-level taxa on basis of overall similarities, instead of on synapomorphies. Our interpretation of Catantopidae in the present paper, as defined by the presence of prosternal process between the forecoxae, is in accordance with Catantopinae of Roberts (1941) and Mistshenko (1952) and Catantopidae of Harz (1975) and is equivalent to the “Spine-breasted Acrididae” as keyed out in Otte (1981). Throughout the paper, we consistently use the name Catantopidae except when discussing its treatment by various previous authors. In the latter case, they were referred to as were originally treated by these authors, such as Catantopinae or Catantopini. The same rule is also consistently applied to other taxa, e.g. Egnatiidae.

About 327 species in 96 genera of Catantopidae (sensu Mistshenko 1952) are known from China, with representatives from both the Palearctic (21.44%) and the Oriental regions (79.56%) (adjusted according to Huang and Chen (1999)). The Chinese fauna of catantopids represents 15% of world genera of the family (data from Vickery and Kevan, 1983) and all of the tribes recognized by Mistshenko (1952) or subfamilies by Dirsh (1975). In this study, we examined a total of 2, 536 specimens representing 240 species in 94 genera, accounting for 73% and 98% of the total number of species and genera known from the country, respectively. Of the 94 examined genera, 84 genera were included in our phylogenetic analysis while the other eight were excluded (Appendix 1). The reasons for the exclusion are: 1) type specimens were not available for examination and no other specimen of these genera had been collected since the original publications, such as Tzacris Tinkham and Chapacris Tinkham; 2) only females were then discovered, such as Liaopodisma Zheng. In addition, we also left out several genera that were described after the data collection stage of this study, such as Caryandoides (Liu and Li 1995, Özdikmen 2009) and Tectiacris (Wei and Zheng 2005). The final inclusion of taxa represented all of the tribes recognized by Mistshenko (1952) and subfamilies by Dirsh (1975).

The majority of the study materials of the present project were provided by the following institutions (curators in parentheses):

Entomological Museum, Shaanxi Normal University, Xi’an, Shaanxi Province (Shengquan Xu)

Entomological Museum, Zhongshan University, Guangzhou, Guangdong (Geqiao Liang)

Entomological Museum, Research Institute of Entomology, Chinese Academy of Sciences, Shanghai (Kailing Xia)

Entomological Museum, Beijing Institute of Zoology, Chinese Academy of Sciences, Beijing (Chunmei Huang)

Zoological Museum, Northwest Plateau Institute of Biology, Chinese Academy of Sciences, Xining, Qinghai (Xiangchu Yin)

Because of the lack of phylogenetic analysis of Acridoidea at levels above subfamily, we had to rely on previous systematic studies on Acridoidea for outgroup selection. All existing classifications of Acridoidea treated Catantopinae, Egnatiinae, Acridinae, and Oedipodinae as being closer to each other than they are to Pyrmorphinae and Pamphaginae (Roberts 1941, Mistshenko 1952, Dirsh 1956, 1961, 1975; Yin 1982, Xia 1994). Dirsh (1961, 1975) suggested that Egnatiinae was closer to Catantopinae than any other subfamily of his Catantopidae because Egnatiinae possesses a Comstock-Kelogg gland, which is otherwise believed to occur only in Catantopinae. Furthermore, Egnatiinae and Catantopinae share similar folds and sculpture in the internal surface of foregut, which are different from those of Oedipodinae. Stebaev et al. (1984) also agreed on a close relationship between Egnatiinae and Catantopinae on basis of cytogenetical, taxonomical and ecological data, but considered the Egnatiinae as a tribe within Catantopinae. Many contemporary acridologists are in agreement about a close relationship between Egnatiinae and Catantopinae (e.g., David Hollis, pers. comm.). Because of the close relationship between Egnatiidae and Catantopidae, very likely as sister clades, and the lack in Egnatiidae of the obvious catantopid synapomorphy of having a prosternal process between the forecoxae, the family Egnatiidae represented by the two genera Egnatius Stal and Egnatioides Voss, was used as outgroup for the phylogenetic analysis of Catantopidae relationships.

Terms and abbreviations used in the present study followed B.-Bienko and Mistshenko (1951) for external morphology and Dirsh (1956, 1961, 1975) for genitalia structures.

Specimens for the study were selected in the following order of priority: 1) type specimens, 2) specimens determined by the author of the taxon, and 3) specimens determined by experts of the taxon. All characters were coded from direct observation of specimens, except in a few instances where characters of a species were coded based on illustrations and descriptions from monographs or reviews (Willemse 1956, 1957; B-Bienko and Mistshenko 1951, Mistshenko 1952, Hollis 1975).

External morphology was surveyed before specimens were dissected for examination of genitalia characters. When available, multiple individuals were examined for each species and multiple species for each genus. For polymorphism at species level, we took an approach similar to, but much more restricted than, the “majority state rule” proposed by Wiens (1995). We generally avoided characters that are polymorphic at species level, and only in very few cases, coded species in question as the predominant state when the other state(s) was rare (presence rate < 15%). In a few cases when character polymorphism occurred at generic level, the characters in question were initially coded as missing for the genus, but were eventually abandoned and not included in the analysis. Some of the characters with three or more states were treated prior to the cladistic analysis as ordered or additive characters, i.e., the transformation series was hypothesized to be 0-1-2 and so on. This was done only when it was possible to order the states unambiguously, e.g., for measurement ratios, and ordered characters are indicated in Appendix 1. In a few cases, one of the states of a main character was more finely subdivided into one or two subsidiary character(s). Taxa with other states of the main character were coded as having state unknown (character not applicable) in the subsidiary character. This commonly used method has been referred to as ‘state-unknown coding’ (Nordlander et al. 1996). The method may give incorrect lengths for some trees when there is homoplasy in the main character and different subsidiary states are ancestral for the different clades having the subdivided state of the main character (Maddison 1993). It has been suggested to use step matrices to represent main/subsidiary character systems exactly (Maddison and Maddison 1992), but this will slow down calculations considerably and is especially impractical for relatively large matrixes like ours. In the present study, therefore, we consistently used state-unknown coding for main/subsidiary character systems and weighted all main and subsidiary characters equally. More detailed discussion about the application of the method is found in Nordlander et al. (1996).

The final matrix contained 87 terminals, including outgroup and 86 catantopid genera in the ingroup, and 88 characters, of which 79 were phylogeny-informative and the other nine were autapomorphies (Appendix 2-3). The autapomorphic characters were excluded from the final cladistic analyses and not counted when calculating tree length, CI, or RI. Nonetheless, they were kept in the matrix for their taxonomic values and potential use in future phylogenetic studies involving the included taxa.

PAUP version 4.0 beta10 (Swofford 2003) was used for phylogenetic analyses. The large number of taxa and characters included in this study did not allow the use of exact searching algorithms. Therefore, we used a combination of several ‘shortcut’ approaches. We first used PAUPRat (Sikes and Lewis 2001) to generate batch files for maximum parsimony analysis within PAUP using the Parsimony Ratchet method described by Nixon (1999). We performed 30 repetitions of the Parsimony Ratchet analysis, with 200 iterations per run as suggested by Sikes and Lewis (2001), giving a total of 6, 000 iterations. The single shortest tree from each of the 6, 000 iterations were then loaded into computer memory for comparison and only the shortest trees over all iterations were kept and duplicates of trees were removed. Because these overall shortest trees were each only the single best tree retained from a particular iteration, they each were probably one of the many possible equally most parsimonious trees or one of the less than most parsimonious trees that actually exist for the dataset. Therefore, these trees were further subjected to TBR branch swapping in order to find out all possible trees of equal or shorter length. To ensure that we find the best trees, we also analyzed our dataset in NONA 2.0 (Goloboff 1999a), a program said to be much faster than competitors like PAUP (Goloboff 1999b). For NONA analyses, we started with MULT*50 (randomize order of taxa, create a weighted Wagner tree, swap using TBR, and with 50 replications) and then swapped the shortest trees from MULT analysis using Max*, which is equal to PAUP’s TBR swapping. NONA was also used for calculation of Bremer Support values (/decay index) for branches (Bremer 1994) while PAUP was used for diagnosis of apomorphic characters supporting each branch. TNT (Goloboff et al. 2011), a program that implemented the tree search methods of NONA as well as additional search methods, including sectorial search, tree drifting, and tree fusing (Goloboff 1999b), was also used for Parsimony Ratchet analysis of the dataset with options comparable to afore-described NONA analysis. The other so-called “New Technology” searching techniques were also used with default options of the software, but were not extensively explored because our dataset was not too large and thus further aggressive approximation was not considered necessary.

We examined a total of 116 characters, including 84 characters of external morphology and 32 characters of male genitalia morphology. Twenty-eight characters were excluded from out analysis either because they were too variable across examined species of a genus to reach a generic consensus or because they were continuous and discrete coding of character states was impossible. In addition, characters of body color patterns, although important for identification of some species of the family, were found to be too variable, both among individuals of species and among species of genera, to be of much use in resolving phylogenetic relationships within Catantopidae and were therefore excluded from the present study. The eighty-eight characters included in the final character matrix consist of 71 external morphological characters and 17 genitalia characters (Appendix 2). Character fit on the shortest trees, as expressed by the consistency index (CI) and retention index (RI), was lower for characters of male genitalia morphology in comparison to characters of external morphology (Table 2).

Using maximum parsimony analysis with Nixon’s ratchet method, we found in thirteen of our 30 replications and 218 of the 6, 000 iterations a tree with the shortest length of 688 steps (L=688, CI = 0.17, RI = 0.55). With duplicate trees deleted, the final number of the shortest trees was 204;subsequent swapping of these optimal trees using TBR did not find shorter trees, but found a total of 22, 354 equally most parsimonious trees. Figs. 1–2 and Fig. 3 show the strict consensus tree with Bremer Support for completely resolved branches and the 50% majority consensus tree with percentage of branches appearing in all shortest trees summarized, respectively.

Searching with NONA 2.0 (Hold=10, 000–30, 000, Mult*50, and Max*) did not find trees shorter than those found with PAUP 4.0 using parsimony ratchet method. Although we were always able to find trees of the shortest length in a few minutes with NONA, our searches invariably resulted in only about 50 trees with MAX*, even when we increased the number of trees to be held in memory to 30, 000. Further swapping using SSWAP*2 and MSWAP*2 apparently would take a long time (3.2GHZ CPU frequency and 1G RAM) and were terminated after a few hours. Comparison of the NONA trees with PAUP trees showed that they were a (small) subset of the trees we found using ratchet method in PAUP. Searching with TNT, either ratchet method or other new technology methods, did not resulted in shorter trees.

Although the number of shortest trees found by our cladistic analyses is huge, the phylogenetic relationship among genera at the base of the cladogram was well resolved, and all basal clades were also relatively well supported with Bremer Support values ranging mostly from 3 to 8 (Fig. 1). The majority of genera, 71 out of 88, fell into the monophyletic Clade A (Fig. 1), which is a polytomy consisting of several relatively well-supported monophyletic clades (Fig. 2: A, A1–5; clade A3 is only supported by a Bremer Support value of 1) as well as a number of unresolved genera / pairs of genera (Fig. 2: A). When a 50% majority consensus tree was calculated, better resolution within Clade A is achieved (Fig. 3, A, B2–B6). In comparison to the strict consensus tree, a sister relationship between A1 and the rest of the clade is supported by 99% of all shortest trees (Fig. 3: A), and A5 (Fig. 2: A5) is supported as the sister clade of the clade consisting of the rest of the genera with improved within-clade resolution (Fig. 3: B5), and (Fruhstorferiola + Tonkinacris) becomes the sister clade to the clade including all members of Clade A except clade A1 and B5 (Fig. 3: A). This terminal clade, while supported by 59% of all shortest trees, form a polytomy consisting of several monophyletic, relatively well resolved clades, 12 distinct genera, and three genera pairs. In addition, there is also an increased resolution at the base of Clade A -- B2 consists of A2 and (Dericorys + Spathosternum) (Fig. 3: B2), B3 includes A3 and Bannacris, and an additional clade is resolved (Fig. 3: B6).

Male genital morphology received special attention from Dirsh (1956, 1961, and 1975) when the author established his classification of acridoids. In fact, the various versions of Dirsh classification depended heavily on the male genitalia morphology, and the practice has greatly influenced later systematists of grasshoppers and other orthopteran insects (Hollis 1975, Yin 1982, Ronderos and Cigliano 1991, Xia 1994, Zheng and Xia 1998). Our result showed that character fit on the shortest trees, as expressed by the consistency index (CI) and retention index (RI), was actually lower for characters of male genitalia morphology in comparison to external morphology characters (Table 2), suggesting that genital characters are not as phylogeny informative as previously thought, at least at higher level, and earlier classification systems of grasshoppers in general and Catantopidae in particular probably include many groups that are not natural due to the heavy reliance on genital features. Eberhard (1985) argued that the species-specific diagnostibility of male genitalia is a reflection of both the rate and extent to which they diverge, and any structure so useful taxonomically must have evolved rapidly. In fact, a recent study showed that morphologically very similar species of Melanoplus grasshoppers differ in the shape of the male genitalia and this is probably due to extremely rapid speciation caused by glacial cycles during the Pleistocene glaciations (Knowles and Otte 2000). The rapid evolution of male genitalia morphology is considered to be caused by strong sexual selection on the male imposed by the females (Eberhard 19985, Knowles and Otte 2000). Regardless of the mechanism, male genital features, while very useful in species identification, show high degree of homoplasy and are therefore of limited value in phylogenetic studies, especially at higher levels. Consequently earlier classifications of acridoids as well as Catantopidae need to be revised critically in the light of phylogenetic analyses based on a broad range of characters.

An earlier attempt to study the phylogenetic relation within Catantopidae from Australia found almost no resolution, especially at the base (Key and Colless 1993), which is strikingly different from the results of our study where the phylogenetic relationship was reasonably resolved, especially at the base. Key and Colless (1993) was able to assemble an impressive dataset consisted of 166 terminals and 104 characters, but unfortunately provided otherwise very limited information about their dataset, which prevents us from interpreting exactly why there is such a big difference between their results and ours. Several factors might have contributed to this. For example, their study is based on males only. While male characteristics are frequently the only useful features for species identification, especially for closely related species, males of different grasshopper species may have been subjected to sexual selection and developed convergent similarities similar to what we have discussed earlier about male genital characteristics. In addition, the authors only used Neighbor-joining and Wagner parsimony without further branch swappingin their analyses, and it is thus very unlikely that what the authors found were the shortest trees. It would be of interest to request from the authors their dataset and reanalyze it using the currently available computation power that is far more superior than it was almost two decades ago. Computation power is especially relevant for analyzing dataset of their size.

Rowell and Flook (1998) presented a phylogenetic tree for the Acridoidea based on the mitochondrial DNA sequences 12S and 16S. They found support for several catantopid clades, i.e., Oxyinae, Podisminae+Melanoplinae, and Coptacridinae. In addition, their study also supported as monophyletic the clade consisting of Cyrtacanthacridinae, Calliptaminae, Catantopinae s. str., and Eyprepocnemidinae. These clades are mostly supported by the present study except the monophyly of (Cyrtacanthacridinae + Calliptaminae + Catantopinae s.s. + Eyprepocnemidinae). While a sister relationship between Cyrtacanthacridinae and Calliptaminae is supported by the present study, Catantopinae is supported as a monophyletic basal clade in the family cladogram and Eyprepocnemis as a member of Calliptaminae (Fig. 4).

Rowell and Flook (1998) also suggested that the Acridoidea ‘seems to be the product of a single explosive radiation’ because they were unable to find a resolution at the subfamily level for the basal acridoids. However, this conclusion, according to the authors, is based on a ‘preliminary analysis’, for which the method was not described, and therefore has to be treated with caution. Meanwhile, the result of the study may be biased simply because of the used genes being inadequate with regard to the divergence level and evolution rate of the study group. According to Simon et al. (1994), the mitochondrial rRNA genes of 12S and 16S are considered to be mostly useful at the population level where highly variable sites have not yet experienced multiple substitutions and at deep levels of divergence where the more conserved sites of these genes supply useful phylogenetic information. At intermediate levels of divergence, however, the relatively variable sites probably have experienced multiple substitutions that may obscure phylogenetic signals. In addition, the rates and patterns of evolution of mitochondrial rRNA genes can vary greatly among taxa (Simon et al. 1994, and references therein). The particular analysis of Rowell and Flook (1998) of Acridoidea based on these two genes might just deal with this ‘intermediate level of divergence’ for the Orthoptera. It would be interesting to reanalyze their dataset to resolve the phylogenetic relationship at various levels with in the superfamily, e.g., to include all their major lineages, but include only a few of their sampled species for each of these lineages, or alternatively, analyze each of these major lineages with all their sampled species included. Unfortunately, the article provided neither the sequences nor genbank accession numbers for the sequences.

To our knowledge, the present study is the most comprehensive of its kind to study the higher level phylogeny of orthopteran insects in terms of the number of taxa sampled and characters examined and coded. Through this study we were able to demonstrate that the external morphology of orthopteran insects can be a very useful source for assessing higher-level phylogeny. For example, the study provided complete resolution for the basal relationships of the Catantopidae (Fig. 1), Nonetheless, our dataset were unable to provide an unambiguous solution for the relationships within the largest terminal clade that comprise 80% of all sampled genera in this study (Figs. 2, 3).It is generally accepted that phylogenetic hypotheses basing on as many independent lines of evidence as possible have the highest explanation value (Nixon and Carpenter 1996a). Combining morphological and molecular data should be the direction for future phylogenetic studies of orthopteran insects including Catantopidae. In addition, our study sampled only taxa from China, which was necessary due to the lack of resources, and future phylogenetic studies of Catantopidae should include representative taxa from other areas of the world.

Based on the strict consensus tree and the 50% Majority-rule consensus of the 22, 355 shortest trees, we hereby outline a scheme for the classification for the family Catantopidae from China. As we discussed above, a comprehensive phylogenetic study based on a more inclusive taxon sampling from all regions of the world and including both morphology and molecular sequences is needed for highly resolving the phylogenetic relationship within the family, especially with regard to the relationship between and within the subfamilies Coptacridinae, Oxyinae, and especially Podisminae (see below). Therefore, the purpose of our outline is to serve as a basis for further studies, rather than as formal classification.

According to this scheme, the Chinese Catantopidae can be classified into seven subfamilies: Habrocneminae, Catantopinae, Cyrtacanthacrinae, Calliptaminae, Coptacridinae, Oxyinae, and Podisminae (Fig. 4). Among the seven recongnized subfamilies, Habrocneminae, Catantopinae, Cyrtacanthacrinae, and Calliptaminae are unambiguously supported as monophyletic clades, and the relationship of each to the rest of the family are completely resolved (Fig. 1, Fig. 4). Coptacridinae and Oxyinae, although each relatively well supported as monophyletic clade, are part of a crown clade that is highly unresolved in terms of within clade relationship (Clade A, Fig. 2). The monophyly of Podisminae, and the resolution of its relationship with Coptacridinae and Oxyinae are only supported by the 50% Majority-rule consensus, which is considered as a compromised solution in phylogenetic systematics (Nixon and Carpenter 1996b). Our analyses also identified within the subfamily Podisminae five monophyletic clades (Fig. 4), which may be treated as tribes: Fruhstorferiolini, Promeosternini, Dericorythini, Traulini, and Podismini. Finally, the rest of the genera within Podisminae are temporarily lumped together in the tribe ‘Melanoplini’ for convenience until further phylogenetic information becomes available.